HYPOTHESIS AND THEORY article
Front. Psychiatry
Sec. Computational Psychiatry
This article is part of the Research TopicArtificial Intelligence in Mental Health Care: Toward Human-Centered and Clinically Grounded InnovationView all articles
Perceive-Assess-Dose-Safeguard (PAD-S): A Safety-Gated State-Action Grammar for Psychotherapy Micro-Decisions in Computational Psychiatry
Provisionally accepted
- Kliniken Erlabrunn gGmbH, Breitenbrunn, Germany
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
Psychotherapy unfolds as a sequence of rapid micro-decisions under uncertainty. Within seconds, clinicians integrate verbal, paraverbal, embodied, and relational cues, estimate the patient's momentary capacity for affective work, choose an intervention dose, and apply stop rules to prevent overwhelm and rupture. Computational psychiatry offers principled frameworks for sequential decision-making, but progress in computational psychotherapy remains constrained by the lack of clinically grounded, machine-readable grammars that capture therapist micro-decisions in context. I introduce the Perceive-Assess-Dose-Safeguard (PAD-S) decision matrix as a safety-gated state–action grammar for psychotherapy micro-decisions. PAD-S formalizes four "front-of-system" signals—defensive/avoidant organization (DEF), anxiety/arousal and tolerance (ANX), patient progression toward direct experience and action (PRO), and self-attack/shame processes (SUP)—together with three safety thresholds (A–C) that gate intervention dose. Each decision point can be logged as an "episode line" (trigger, state, threshold, action, and expected functional impact), enabling transcript annotation and structured datasets. PAD-S is grounded in experiential dynamic psychotherapy (EDT/ISTDP) yet expressed as an orientation-translatable representation layer: DEF can be read as avoidance/safety behavior, ANX as arousal/tolerance, PRO as approach and value-consistent action, and SUP as self-criticism/shame. I show how PAD-S trajectories can interface with hybrid neural–cognitive models such as SPICE to discover sparse, interpretable equations of process change, and I outline testable hypotheses and feasible pilot studies (reliability, outcome linkage, and modeling) to evaluate the framework.
Keywords: active inference, Computational Psychiatry, Computational psychotherapy, human-in-the-loop, Interpretable AI, Mini-ICF-APP, psychotherapy process coding, Spice
Received: 18 Nov 2025; Accepted: 11 Feb 2026.
Copyright: © 2026 Niederlohmann. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence: Eik Niederlohmann
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.